Many neuropsychiatric disorders of childhood may reflect subtle deviations in brain development. By acquiring and analyzing MRI images from several clinical groups and healthy controls in an identical fashion we have been able to address important issues of specificity. In the process of providing valid control subjects we have acquired one of the world's largest cerebral MRI data sets of well-screened healthy children and adolescents. The enormous variability of brain structure sizes highlights the need for large sample sizes and longitudinal study designs, which are underway. Sexually dimoprhic patterns of normal brain development may shed light on sex differences in age of onset, prevalence, and symptomatology noted in most childhood neuropsychiatric disorders. For instance, the relatively larger size of the caudate nucleus in females may be related to the male-greater-than-female prevalence of ADHD and Tourette's disorder, which are both associated with decreased basal ganglia size. Also, the male-only robust adolescent increases in lateral ventricular volume may be related to the higher male incidence of adolescent onset schizophrenia. Our new image analysis technique now allows us to segment white matter, gray matter, and cerebral spinal fluid on brain MR images and reveals a robust increase with age in white matter, but not gray matter, for our age range of 4 to 18 years. A PET O15 study of normal readers supported a "single-route" connectionist model of reading in which both real and pseuodowords are processed by similar mechanisms in a common neural network. Dyslexics differ from controls in posterior bilateral language regions regardless of the localizing value of the task.